[Intel-gfx] [PATCH i-g-t v7] tests/perf_pmu: Improve accuracy by waiting on spinner to start

Chris Wilson chris at chris-wilson.co.uk
Mon Mar 19 14:02:38 UTC 2018


Quoting Tvrtko Ursulin (2018-03-19 13:56:05)
> @@ -146,7 +198,7 @@ static int emit_recursive_batch(igt_spin_t *spin,
>  
>         /* recurse */
>         fill_reloc(&relocs[obj[BATCH].relocation_count],
> -                  obj[BATCH].handle, (batch - spin->batch) + 1,
> +                  obj[BATCH].handle, (batch - batch_start) + 1,

D'oh.

>                    I915_GEM_DOMAIN_COMMAND, 0);
>         if (gen >= 8) {
>                 *batch++ = MI_BATCH_BUFFER_START | 1 << 8 | 1;
> @@ -167,18 +219,23 @@ static int emit_recursive_batch(igt_spin_t *spin,
>         obj[BATCH].relocation_count++;
>         obj[BATCH].relocs_ptr = to_user_pointer(relocs);
>  
> -       execbuf.buffers_ptr = to_user_pointer(obj + (2 - execbuf.buffer_count));
> -       execbuf.rsvd1 = ctx;
> +       execbuf->buffers_ptr = to_user_pointer(obj +
> +                                              (2 - execbuf->buffer_count));
> +       execbuf->rsvd1 = ctx;
>  
> -       if (out_fence)
> -               execbuf.flags |= I915_EXEC_FENCE_OUT;
> +       if (flags & OUT_FENCE)
> +               execbuf->flags |= I915_EXEC_FENCE_OUT;
>  
>         for (i = 0; i < nengine; i++) {
> -               execbuf.flags &= ~ENGINE_MASK;
> -               execbuf.flags |= engines[i];
> -               gem_execbuf_wr(fd, &execbuf);
> -               if (out_fence) {
> -                       int _fd = execbuf.rsvd2 >> 32;
> +               execbuf->flags &= ~ENGINE_MASK;
> +               execbuf->flags |= engines[i];
> +
> +               if (flags & POLL_RUN)
> +                       igt_require(gem_can_store_dword(fd, execbuf->flags));

Hmm, we need to lift this. If we abort on a secondary engine, we haven't
yet attached ourselves into the igt_terminate_spin_batches and so we'll
get GPU hangs.
-Chris


More information about the Intel-gfx mailing list